CN112256975A

CN112256975A - Information pushing method and device, computer equipment and storage medium

Info

Publication number: CN112256975A
Application number: CN202011268242.0A
Authority: CN
Inventors: 周菲
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-11-13
Filing date: 2020-11-13
Publication date: 2021-01-22

Abstract

The application provides an information pushing method and device, computer equipment and a storage medium, and belongs to the technical field of internet. The method comprises the following steps: acquiring characteristic information of a first user exchanging a target service; mapping the characteristic information to a target vector space to obtain a characteristic vector; respectively inputting the feature vectors into at least two classification models, classifying the first user by the at least two classification models, and outputting at least two classification results; and in response to the at least one classification result indicating that the probability that the first user is the target user is greater than the target probability, pushing target information to the first user, wherein the target information is used for indicating the redemption of the preferential content of the target service. According to the scheme, the target information comprising the preferential content is pushed to the first user determined as the target user, so that part of the first users can be prompted to continue to exchange the target service through the target information, the total number of users exchanging the target service is increased, and the user loss is avoided.

Description

Information pushing method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of internet technologies, and in particular, to an information pushing method and apparatus, a computer device, and a storage medium.

Background

With the development of internet technology, users can enjoy various services provided by internet service providers, such as music playing services, video playing services, game acceleration services, cloud storage services, and the like, by paying a certain fee. How to avoid the loss of users and improve the total number of users purchasing services is a problem to be solved by internet service providers.

Currently, in order to maintain user stickiness and increase the number of users purchasing services, a method is generally adopted to send preferential information, such as discount information, bonus information, discount information and the like, to the users according to business experiences of activity operators and activity contents, so as to attract users with expired services to continuously purchase services.

The above scheme has the problems that due to the uneven levels of the activity operators and the incomplete activity content, the scheme is not sustainable and has high randomness, and the effect of the implemented scheme cannot be accurately expected, so that the risk of user loss exists.

Disclosure of Invention

The embodiment of the application provides an information pushing method, an information pushing device, computer equipment and a storage medium, and by pushing target information comprising preferential content to a target user, the target information can prompt part of the target users to continuously purchase target services, so that the total number of users purchasing the target services is increased, and the loss of the users is avoided. The technical scheme is as follows:

in one aspect, an information pushing method is provided, where the method includes:

acquiring characteristic information of a first user who exchanges a target service, wherein the characteristic information is used for indicating the use characteristic and the exchange characteristic of the first user to the target service;

mapping the characteristic information to a target vector space to obtain a characteristic vector;

respectively inputting the feature vectors into at least two classification models, classifying the first user by the at least two classification models, and outputting at least two classification results, wherein the classification results are used for indicating the probability that the first user is a target user, and the target user is a user who cannot redeem the target service again;

and in response to the fact that the probability that the first user is a target user is indicated by at least one classification result to be larger than the target probability, target information is pushed to the first user, and the target information is used for indicating to redeem the preferential content of the target service.

In another aspect, an information pushing apparatus is provided, the apparatus including:

the information acquisition module is used for acquiring characteristic information of a first user who exchanges a target service, wherein the characteristic information is used for indicating the use characteristic and the exchange characteristic of the first user on the target service;

the information processing module is used for mapping the characteristic information to a target vector space to obtain a characteristic vector;

the classification module is used for respectively inputting the feature vectors into at least two classification models, classifying the first user by the at least two classification models and outputting at least two classification results, wherein the classification results are used for indicating the probability that the first user is a target user, and the target user is a user who cannot redeem the target service again;

and the information pushing module is used for responding to the fact that the probability that the first user is the target user is greater than the target probability in response to the at least one classification result, pushing target information to the first user, wherein the target information is used for indicating the exchange of the preferential content of the target service.

In an alternative implementation, the training of the at least two classification models includes:

acquiring a positive sample and a negative sample, wherein the positive sample is a user which is historically exchanged and is currently exchanged for the target service, and the negative sample is a user which is historically exchanged and is not currently exchanged for the target service;

obtaining sample characteristic information of the positive sample and the negative sample, wherein the sample characteristic information is used for indicating the use characteristic and the redemption characteristic of the positive sample and the negative sample on the target service;

and training to obtain the at least two classification models based on the sample characteristic information.

In an alternative implementation, the obtaining positive and negative samples includes:

acquiring a target date, wherein the target date and the current date are separated by a third time length;

according to the target date, acquiring a plurality of users who have historically redeemed the target service for not less than a fourth time and have the target service expired on the target date;

selecting users meeting a first condition from the plurality of users as the positive sample, wherein the first condition is that the target service is redeemed for not less than a fifth time within a target time range corresponding to the target date;

selecting users satisfying a second condition from the plurality of users as the negative sample, wherein the second condition is that the target service is not redeemed within a target time range corresponding to the target date.

In an optional implementation manner, the training to obtain the at least two classification models based on the sample feature information includes:

inputting the sample characteristic information of any positive sample or negative sample into a logistic regression model corresponding to the iteration process, and determining a loss function of the logistic regression model;

regularizing the loss function, and adjusting parameters of the logistic regression model according to the regularized loss function;

in response to the logistic regression model satisfying a training end condition, treating the logistic regression model as one of the at least two classification models.

In an optional implementation manner, the inputting the sample feature information of any positive sample or negative sample into the logistic regression model corresponding to the current iteration process includes:

and normalizing the sample characteristic information of any positive sample or negative sample to obtain normalized characteristics, and inputting the normalized characteristics into a logistic regression model corresponding to the iteration process.

and mapping the sample characteristic information of any positive sample or negative sample into a single-hot coded vector, and inputting the single-hot coded vector into the logistic regression model corresponding to the iteration process.

In an alternative implementation, the target service is an acceleration service;

the information acquisition module is used for acquiring at least two of an acceleration time characteristic, an acceleration delay packet loss characteristic, a login characteristic, an acceleration content characteristic, a service switching behavior characteristic used for indicating to use other acceleration services, a first exchange behavior characteristic used for indicating to exchange the acceleration service recently and a second exchange behavior characteristic used for indicating to exchange the acceleration service historically of a first user who exchanged the target service; and performing data preprocessing on the acquired at least two characteristics to obtain the characteristic information of the first user.

In an optional implementation manner, the information pushing module is configured to determine at least two probabilities according to the at least two classification results; in response to any probability of the at least two probabilities being greater than the target probability, pushing the target information to the first user; or, in response to that the at least two probabilities are both greater than the target probability, pushing the target information to the first user; or, in response to the probability exceeding a target ratio in the at least two probabilities being greater than the target probability, pushing the target information to the first user.

In an optional implementation, the apparatus further includes:

and the user acquisition module is used for acquiring a user meeting a target condition as the first user, wherein the target condition is that the target service expires after a first time length, and the time length for purchasing the target service is not less than a second time length.

In an optional implementation manner, the information processing module is configured to map the feature information into a one-hot coded vector, and use the one-hot coded vector as the feature vector.

In another aspect, a computer device is provided, where the computer device includes a processor and a memory, where the memory is used to store at least one program code, and the at least one program code is loaded and executed by the processor to implement the operations performed in the information pushing method in the embodiments of the present application.

In another aspect, a computer-readable storage medium is provided, where at least one program code is stored, and the at least one program code is loaded and executed by a processor to implement the operations performed in the information pushing method in the embodiments of the present application.

In another aspect, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer program code, the computer program code being stored in a computer readable storage medium. The processor of the computer device reads the computer program code from the computer-readable storage medium, and the processor executes the computer program code, so that the computer device performs the information pushing method provided in the above-described aspects or various alternative implementations of the aspects.

The technical scheme provided by the embodiment of the application has the following beneficial effects:

the embodiment of the application provides an information pushing method, which includes the steps that characteristics beneficial to classifying first users can be obtained by obtaining usage characteristics and exchange characteristics of the first users exchanging target services to the target services, then the first users are classified according to characteristic vectors corresponding to the characteristics based on at least two classification models, the probability of whether the first users are target users incapable of continuously exchanging the target services can be determined, target information including preferential content is pushed to the first users based on the relation between the probability and the target probability, the target information can prompt the first users to continuously exchange the target services, the total number of users exchanging the target services is increased, and user loss is avoided.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic diagram of an implementation environment of an information pushing method provided according to an embodiment of the present application;

fig. 2 is a flowchart of an information pushing method provided according to an embodiment of the present application;

fig. 3 is a flowchart of an information pushing method provided according to an embodiment of the present application;

fig. 4 is a schematic architecture diagram of an information push system according to an embodiment of the present application;

FIG. 5 is a flow chart of another information pushing method provided according to an embodiment of the present application;

FIG. 6 is a block diagram of an information pushing apparatus provided according to an embodiment of the present application;

fig. 7 is a block diagram of a terminal according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of a server provided according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

The following description refers to the terminology used in the examples of the present application.

Line sampling and column sampling are common modes in integration models such as random forests, XGboost and the like. Its main function is to speed up training and prevent overfitting.

The grid search method is an exhaustive search method for specifying parameter values, and an optimal learning algorithm is obtained by optimizing parameters of an estimation function through a cross validation method. Namely, the possible values of each parameter are arranged and combined, and all the possible combination results are listed to generate a 'grid'. Each combination was then used for SVM (Support Vector Machine) training and performance was evaluated using cross-validation. After all parameter combinations are tried by the fitting function, a proper classifier is returned, the optimal parameter combination is automatically adjusted, and parameter values can be obtained through clf.

The voting mechanism is divided into soft voting and hard voting, and the principle adopts the idea of majority-obeying minority. Hard voting: and (4) directly voting a plurality of models, wherein the class with the largest final vote number is the final predicted class. Soft voting: the method is the same as a hard voting principle, the function of setting the weight is added, different weights can be set for different models, and then different importance degrees of the models are distinguished.

hive is a data warehouse tool based on Hadoop, which is used for data extraction, transformation and loading, and is a mechanism capable of storing, querying and analyzing large-scale data stored in Hadoop. The hive data warehouse tool can map the structured data file into a database table, provide SQL query function and convert SQL sentences into MapReduce tasks for execution. Hive has the advantages that the learning cost is low, rapid MapReduce statistics can be realized through similar SQL sentences, MapReduce is simpler, and a special MapReduce application program does not need to be developed. hive is well suited for statistical analysis of data warehouses.

Hereinafter, an implementation environment of the information push method provided by the embodiment of the present application is described. Fig. 1 is a schematic diagram of an implementation environment of an information push method according to an embodiment of the present application. Referring to fig. 1, the implementation environment includes a terminal 101 and a server 102.

The terminal 101 and the server 102 can be directly or indirectly connected through wired or wireless communication, and the application is not limited herein.

Optionally, the terminal 101 is a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, or the like, but is not limited thereto. The terminal 101 is installed and operated with an application program. The application program is any one of a music playing application program, a video playing application program, a game acceleration application program and a cloud storage application program. Illustratively, the terminal 101 is a terminal used by a user, and the user uses any one of a music playing service, a video playing service, a game acceleration service, and a cloud storage service provided by an internet server provider, using the above-described application installed and run in the terminal 101.

Optionally, the server 102 is an independent physical server, may also be a server cluster or a distributed system formed by a plurality of physical servers, and may also be a cloud server providing basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a Network service, cloud communication, a middleware service, a domain name service, a security service, a CDN (Content Delivery Network), a big data and artificial intelligence platform, and the like. The server 102 is used to provide background services for the application. Optionally, the server 102 undertakes primary computational work and the terminal 101 undertakes secondary computational work; or, the server 102 undertakes the secondary computing work, and the terminal 101 undertakes the primary computing work; alternatively, the server 102 and the terminal 101 perform cooperative computing by using a distributed computing architecture.

Optionally, the terminal 101 generally refers to one of a plurality of terminals, and this embodiment is only illustrated by the terminal 101. Those skilled in the art will appreciate that the number of terminals 101 described above may be greater or fewer. For example, the number of the terminals 101 may be only one, or the number of the terminals 101 may be several tens or hundreds, or more. The number of terminals and the type of the device are not limited in the embodiments of the present application.

Optionally, the wireless network or wired network described above uses standard communication techniques and/or protocols. The Network is typically the Internet, but can be any Network including, but not limited to, a Local Area Network (LAN), a Metropolitan Area Network (MAN), a Wide Area Network (WAN), a mobile, wireline or wireless Network, a private Network, or any combination of virtual private networks. In some embodiments, data exchanged over a network is represented using techniques and/or formats including Hypertext Mark-up Language (HTML), Extensible Markup Language (XML), and the like. All or some of the links can also be encrypted using conventional encryption techniques such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec). In other embodiments, custom and/or dedicated data communication techniques can also be used in place of or in addition to the data communication techniques described above.

Fig. 2 is a flowchart of an information pushing method according to an embodiment of the present application, and as shown in fig. 2, the information pushing method is described in the embodiment of the present application by taking the information pushing method as an example. The information pushing method comprises the following steps:

201. the computer device obtains feature information of a first user who redeems a target service, wherein the feature information is used for indicating the use feature and the redemption feature of the target service of the first user.

In the embodiment of the present application, the target service is any one of services provided by internet service providers, such as a music playing service, a video playing service, a game acceleration service, and a cloud storage service. The user can exchange the target service through a resource, wherein the resource is a virtual resource such as points, tickets, electronic money and the like or a physical resource such as legal currency and the like. Alternatively, the user can effect redemption of the target service by purchasing the target service. For a user who redeems a target service, the computer device can authorize to acquire behavior information and attribute information of the user for using the target service and redeeming the target service, and then the computer device can acquire using characteristics of the user for using the target service and redeeming characteristics of the user for redeeming the target service based on the acquired information. When the target service which is converted by any user is about to expire, the user is the first user.

202. And the computer equipment maps the characteristic information to a target vector space to obtain a characteristic vector.

In this embodiment, the computer device can further process the obtained feature information, and map the feature information into a one-hot code to obtain a feature vector.

203. The computer equipment respectively inputs the feature vectors into at least two classification models, the first user is classified by the at least two classification models, and at least two classification results are output, wherein the classification results are used for indicating the probability that the first user is a target user, and the target user is a user who cannot continuously purchase the target service.

In this embodiment of the application, the computer device may input the obtained feature vectors into at least two classification models, classify the feature information of the first user based on the at least two classification models, and output a classification result by each classification model, so as to obtain at least two classification results. And the classification result output by each classification model can indicate the probability that the first user is the target user. Wherein, the higher the probability is, the higher the probability that the first user will not continuously redeem the target service after the service is expired is indicated; the lower the probability, the lower the probability that the first user will not continue to redeem the target service after the service expires.

204. In response to at least one classification result indicating that the probability that the first user is a target user is greater than a target probability, the computer device pushes target information to the first user, wherein the target information is used for indicating to redeem the preferential content of the target service.

In this embodiment of the application, the computer device may determine whether the first user is the target user according to the obtained at least two classification results, and optionally, when any classification result indicates that the probability that the first user is the target user is greater than the target probability, determine that the first user is the target user; or when all the classification results indicate that the probability that the first user is the target user is greater than the target probability, determining that the first user is the target user; or when the classification result exceeding the target proportion indicates that the probability that the first user is the target user is greater than the target probability, determining that the first user is the target user. The embodiment of the present application does not limit this. After the computer device determines that the first user is the target user, the computer device can push target information for indicating the exchange of the preferential content of the target service to the first user, so that the first user can continue to exchange the target service after receiving the target information, the loss of users is avoided, and the total number of users exchanging the target service is increased.

In the embodiment of the application, an information pushing method is provided, and features beneficial to classifying a first user who redeems a target service can be obtained by obtaining the use features and the redemption features of the target service by the first user, then the first user is classified according to feature vectors corresponding to the features based on at least two classification models, the probability of whether the first user is a target user who cannot continuously redeem the target service can be determined, and target information including preferential content is pushed to the first user based on the relationship between the probability and the target probability, so that the target information can prompt the first user to continuously redeem the target service, the total number of users redeeming the target service is increased, and user loss is avoided.

Fig. 2 illustrates the main steps of the information pushing method provided by the embodiment of the present application, which can be applied to a scenario of pushing various services provided by an internet service provider to a target user, and the following description is given based on an application scenario of pushing a coupon to a user of a game acceleration service as an example.

Fig. 3 is a flowchart of an information pushing method according to an embodiment of the present application, and as shown in fig. 3, the information pushing method is described in the embodiment of the present application by taking an application to a server as an example. The information pushing method comprises the following steps:

when the target service is the game acceleration service, the first user who exchanged the target service is a user who purchased the game acceleration service through virtual money or physical money. Optionally, the user can purchase the game acceleration service for a certain duration, such as one day, seven days, one month, one quarter, one year, etc., through virtual money or physical money according to the actual needs of the user. Accordingly, since the game acceleration service has a certain timeliness, the game acceleration service purchased by the user may expire after a certain period of time, and the first user can purchase the game acceleration service again through virtual money or physical money or no longer purchase the game acceleration service. The method and the device can determine whether the coupon needs to be issued to the user by predicting whether the user will continuously purchase the game acceleration service after the game acceleration service expires, so that the possibility of user renewal is improved. That is, for a user who is predicted not to continue purchasing the game acceleration service after expiration, the possibility that the user continues purchasing the game acceleration service can be increased by pushing the information including the coupon thereto.

In the embodiment of the application, the server can determine whether the user is a target user who does not continuously purchase the target service through the classification model. The steps of the server training the classification model are seen in steps 301 to 303.

301. The server obtains training samples.

In the embodiment of the application, the training samples are divided into positive samples and negative samples, the positive samples are users which have been redeemed historically and are currently redeemed for the target service, and the negative samples are users which have been redeemed historically and are not redeemed for the target service currently.

For example, when the target service is a game acceleration service, the positive sample is a user who has continuously purchased the game acceleration service for a number of days equal to or more than one month and purchased the game acceleration service for a period equal to or more than one month after the game acceleration service expires; the negative sample is the number of days that the game acceleration service is continuously bought for one month or more, and the game acceleration service is not bought for the user until now after the game acceleration service is expired. The user purchases the game acceleration service again seven days before or seven days after the game acceleration service expires, and the user purchases the game acceleration service again more than seven days after the game acceleration service expires, but does not belong to the behavior of purchasing the game acceleration service again after the game acceleration service expires, namely the user with the behavior does not belong to the positive sample user. It should be noted that, the respective time periods related to the above examples are only exemplary illustrations, and the embodiments of the present application do not limit this.

In an alternative implementation, the server, when obtaining the training samples, can obtain based on a target date. The target date is separated from the current date on which the training samples were obtained by the server by a third length of time. Correspondingly, the step of obtaining the positive sample and the negative sample by the server comprises the following steps: the server acquires a target date, and then acquires a plurality of users of which the history is exchanged for the target service with the time length not less than the fourth time and the target service expires on the target date according to the target date; then selecting users meeting a first condition from the plurality of users as a positive sample, wherein the first condition is that target services with not less than a fifth time length are exchanged within a target time range corresponding to a target date; and selecting users meeting a second condition from the plurality of users as a negative sample, wherein the second condition is that the target service is not redeemed within a target time range corresponding to the target date. Wherein the third duration is one or more months. The fourth time period and the fifth time period can be equal or unequal. The target time range corresponding to the target date is three days, five days or seven days before and after the target date respectively. The third time duration, the fourth time duration, the fifth time duration and the target time range are not limited by the application.

302. The server obtains sample characteristic information of the training samples.

In the embodiment of the application, the behavior information and the attribute information of the positive sample and the negative sample can be stored in a data table form, and the server can extract the feature data of the positive sample and the negative sample from the data table, so as to obtain the sample feature information of the training sample.

In the embodiment of the application, when the target service is a game acceleration service, seven types of characteristics, namely an acceleration time characteristic, an acceleration delay packet loss characteristic, a login characteristic, an acceleration content characteristic, a service switching behavior characteristic for indicating use of other acceleration services, a first redemption behavior characteristic for indicating recent redemption of the acceleration service, and a second redemption behavior characteristic for indicating historical redemption acceleration services, are selected as characteristics included in the sample characteristic information. Optionally, the server may further select more types of features, such as a user portrait feature, a history preferential feature, and the like, which is not limited in this embodiment of the application. Of course, when the target service is another type of service, the server can obtain a corresponding plurality of types of features.

For example, the acceleration time characteristic is a behavior characteristic of a user when using the game acceleration service, and is generally counted based on three weeks before the game acceleration service expires. The statistical content comprises: number of days accelerated in first week; number of days accelerated at the end of the first week; a first week acceleration total duration; a first weekend total acceleration period proportion (representing a proportion of the first weekend total acceleration period to the first week total acceleration period); accelerating the total time of the member game in the first week (the member game refers to the game which can be played only after purchasing the game member); the first week average acceleration duration (calculated as the total acceleration duration divided by the number of acceleration days); the total acceleration time length of 0 to 6 points of the first week is in proportion; first week 0 o 'clock to 6 o' clock acceleration days ratio (representing the ratio of first week 0 o 'clock to 6 o' clock acceleration days to total first week acceleration days); number of accelerated days in second week; number of days accelerated on weekend; a second week total acceleration duration; a second weekend total acceleration duration ratio (representing the ratio of the second weekend total acceleration duration to the second week total acceleration duration); accelerating the total time of the member game to account for the ratio (the member game is a game which can be played only after the game members are charged) in the second week; the second week average acceleration duration (calculated as the total acceleration duration divided by the number of acceleration days); the total acceleration time length is in ratio from 0 point to 6 points in the second week; second week 0 to 6 days accelerated ratio (representing the ratio of second week 0 to 6 days accelerated to second week total days accelerated); third week accelerated days; number of days accelerated on weekend third week; total acceleration duration for the third week; third week weekend total acceleration duration to (representing the proportion of the third week weekend total acceleration duration to the third week total acceleration duration); accelerating the total time of the member game in the third week; average acceleration duration for the third week (calculated as total acceleration duration divided by number of acceleration days); the total acceleration time length of 0 to 6 points in the third week accounts for the ratio; acceleration days 0 to 6 on the third week (representing the ratio of acceleration days 0 to 6 on the third week to the total acceleration days on the third week); total acceleration days three weeks; total number of days accelerated on weekends of three weeks; total acceleration duration for three weeks; a total three week weekend acceleration duration ratio (representing the ratio of the total three week weekend acceleration duration to the total three week acceleration duration); the total three-week accelerated membership game duration ratio (total three-week accelerated average duration divided by the number of accelerated days), total three-week accelerated 0-6 accelerated total duration ratio, total three-week accelerated 0-6 accelerated day ratio (representing the ratio of total three-week accelerated 0-6 accelerated days to total three-week accelerated days), total first-week accelerated duration ratio (representing the ratio of total first-week accelerated duration to total three-week accelerated duration ratio), total second-week accelerated duration ratio, total third-week accelerated duration ratio less than one tenth, and average third-week accelerated duration less than one quarter of the total average accelerated duration ratio, of course, the server can also count four weeks or five weeks before the game accelerated service expires, and the counted time period can be 18-24-point time periods in addition to the 0-6-point time period, the embodiment of the present application does not limit this.

For another example, the accelerated delay packet loss feature is usually counted based on the accelerated delay and the packet loss rate when accelerating the member game in the last week. The statistical content comprises: acceleration delay of 0.98 quantile in accelerating member games in the last week; acceleration delay of 0.9 quantile in accelerating member games in the last week; acceleration delay of 0.7 quantile in accelerating member games in the last week; an acceleration delay average value when accelerating the member game for the last week; the packet loss rate is 0.98 quantile when accelerating the member game in the last week; the packet loss rate is 0.9 quantile when accelerating the member game in the last week; the packet loss rate is 0.7 quantile when accelerating the member game in the last week; average value of packet loss rate when accelerating member game in the last week, etc. Of course, the server may also count only the acceleration delay condition and the packet loss rate when the user accelerates the free game, may also count the acceleration delay condition and the packet loss rate when the user accelerates all games, or respectively calculate the acceleration delay condition and the packet loss rate according to different acceleration scenarios, which is not limited in the embodiment of the present application.

As another example, the login characteristics are behavioral characteristics of the user logging into the game acceleration service, and are typically counted based on the last three weeks of the day. The statistical content comprises: the number of days of last week to log in; the number of days of last two weeks of login to the game; the number of different machines used in three weeks, etc.

For example, the service switching behavior feature may be used to instruct the user to use a feature of another acceleration service, such as a feature of a game acceleration service provided by another service provider. The game acceleration service provided by other service providers is a competitive product, and the game acceleration service related to the application is the product. The statistical content comprises: total days of use of the contest in the last week; total days of use of the contest in the last two weeks. Of course, the server can also count the total days of using the competitive products in the last three weeks, the ratio of the total days of using the competitive products in the last one week to the total days of using the product in the last one week, the ratio of the total days of using the competitive products in the last two weeks to the total days of using the product in the last one week, and the like, which is not limited in the embodiment of the present application.

For example, the accelerated content feature is used to indicate to the user the associated features of free games and member games that have been accelerated by the game acceleration service. The statistical content comprises the following steps: an accelerated number of free games; number of accelerated member games; the number of accelerated member games is proportional; accelerating free game days; speeding up the number of member game days, etc. Of course, the server can also count the number of free game acceleration days, the total number of game acceleration days, and the like, which is not limited in the embodiment of the present application.

As another example, the first redemption behavior characteristic is used to indicate a characteristic of a recent redemption of an acceleration service, where typically the user redeems the acceleration service in a manner of purchase. It is common to count the characteristics of the last time the game acceleration service was purchased. The statistical content comprises: how many days of game acceleration service have been purchased continuously (including multiple purchases with no break in between, i.e., purchases again without expiration); the number of days of the game acceleration service purchased this time; whether the purchase is a game acceleration service with more privileges, such as a Super Virtual Inportant Person (SVIP), is purchased at this time, and the SVIP has more privileges than a common VIP (member); whether the game acceleration service is purchased in the automatic charging mode or not; this way of purchasing game acceleration service.

As another example, the second redemption behavior characteristic is used to indicate a characteristic of historical redemption acceleration services, where typically the user redeems the acceleration services in a manner of purchase. It is common to count the activities of purchasing the game acceleration service for the last three months and the total purchasing activities. The statistical content comprises: days within the service validity period in the last three months; total days of the interval between the game acceleration service purchased in the last three months/the number of game acceleration service purchased in the last three months; the number of times of purchasing the game acceleration service at the interval of 0 (continuously purchasing the game acceleration service) in the last three months is in proportion to the total number of times of purchasing the game acceleration service in three months; the number of times that the interval of purchasing game acceleration service in the last three months is greater than 0 is proportional; how many times the game acceleration service has been purchased in the total purchase game acceleration service behavior; the times of purchasing the game acceleration service monthly card in all the game acceleration service purchasing behaviors; the number of times of purchasing the game acceleration service monthly card in the behavior of purchasing the game acceleration service in total is proportional to the total number of times of purchasing the game acceleration service in the behavior of purchasing the game acceleration service in total; the times of purchasing days more than 31 days in all the behaviors of purchasing the game acceleration service; the times of purchasing days more than 31 days in all the behaviors of purchasing the game acceleration service are in proportion; the times of purchasing days less than 31 days in all the behaviors of purchasing the game acceleration service; the times of purchasing days less than 31 days in all the behaviors of purchasing the game acceleration service are compared; the interval of each purchase in the action of totally purchasing the game acceleration service is several days; the number of times of purchasing an interval of 0 (continuously purchasing the game acceleration service) in the behavior of purchasing the game acceleration service in total is in proportion; the purchase interval is greater than 0 times in the action of purchasing the game acceleration service in total, and the like.

It should be noted that the above-mentioned features acquired by the server are numeric features, for example, the range of the value of the duty feature is 0 to 100. Optionally, the training samples that the server can obtain every day are different, for example, the user a accords with the condition that the server obtains the training samples on the ith day, and is obtained as a positive sample or a negative sample by the server, but does not accord with the condition that the server obtains the training samples on the ith-1 day and the ith-1 day; and the user B does not accord with the condition that the server obtains the training sample on the i-1 th day, accords with the condition that the server obtains the training sample on the i th day, and is obtained as a positive sample or a negative sample by the server, wherein i is a positive integer and represents the current day. Optionally, the server may perform model training using sample feature information of a training sample newly obtained every day, or may perform model training using sample feature information of a training sample obtained on the current day and several days before, for example, days i to i-4, so as to increase the number of training samples.

303. The server trains at least two classification models based on the sample feature information.

In the embodiment of the application, a multi-model fusion mode is adopted, namely, an intersection or a union is taken for the classification results of a plurality of classification models, so that the overall accuracy of the classification models is improved. Because the more similar classification models are fused, the poorer the fusion effect is, at least two classification models trained by the server are independent from each other, and the results have no correlation.

In an alternative implementation, the at least two classification models include an extreme gradient boosting model, a random forest model, and a logistic regression model. Of course, the server can also adopt other models, and the application is not limited to this.

Both the XGBoost (eXtreme Gradient Boosting) model and the RF (Random Forest) model are tree models. The RF model can vote on the decision results of multiple decision trees. The XGBoost is also decided by multiple trees together, but unlike the RF model, the learning process of all the trees in the XGBoost is serial, that is, the learning and decision of the following tree need to depend on the result of the preceding tree, and the following tree learns the residual of the conclusion sum of the preceding trees, that is, the input of the following tree is the difference between the real value and the preceding tree. Optionally, in order to accelerate the training speed of the XGBoost model and the RF model and prevent overfitting of the models, the method of column sampling and increasing the second derivative is adopted for training, so that the accuracy of the models can be improved. The overall effect of the XGboost model is optimal, but the accuracy of the RF model is highest, and the two models have certain differences in the prediction effect bias.

It should be noted that, because the tree model divides the samples according to the split points, the training sample feature vectors of the tree model do not need to be normalized, and the processed feature vectors of the digital type can be directly used as the model input.

An LR (Logistic Regression) model is constructed by combining a linear Regression function and a Sigmoid function. When the LR model is trained, the server can train the LR model through multiple iterations, and in one iteration process, the server can input the sample characteristic information of any positive sample or negative sample into the logistic regression model corresponding to the iteration process, so as to determine the loss function of the logistic regression model. Then, the server can perform regularization processing on the loss function, and perform parameter adjustment on the logistic regression model according to the loss function after regularization processing. In response to the logistic regression model satisfying the training end condition, the server can treat the logistic regression model as one of the at least two classification models.

The linear regression function is shown in formula (1), and the logistic regression function is shown in formula (2).

z＝θ₀+θ₁x₁+θ₂x₂+θ₃x₃...+θ_nx_n＝θ^Tx (1)；

Wherein z represents a linear regression value, θ₀Representing the intercept of the regression line, theta₁，θ₂，...，θ_nEach representing the slope, theta, of n regression lines^TRepresenting the slope matrix and x representing the input eigenvector.

Wherein h is_θ(x) Representing logistic regression values, z representing linear regression values, theta^TRepresenting a slope matrixAnd x denotes an input feature vector.

It should be noted that, in order to reduce the model overfitting, the embodiment of the present application can add a regularization term to the loss function of the LR model. The L2 norm regular is added to an LR model in the embodiment of the application, and the L2 norm regular is that the square of the absolute value of theta is directly added to a loss function to serve as a regular term. Accordingly, the loss function of LR is shown in equation (3).

Wherein J (theta) represents loss value, n represents number of samples, and x_iFeature vector, y, representing the ith sample_iTable i sample label, h_θ(x_i) Represents the logistic regression value of the ith sample,

indicating an L2 norm regularization.

It should be noted that the server may further add an L1 norm regularization to the LR model to replace the L2 norm regularization, where the L1 norm regularization refers to directly adding an absolute value of θ to the loss function as a regularization term, and details are not described here.

It should be noted that, since the regularization term of the loss function is the sum of the weights in the linear regression, the weights need to be numbers in the same range, and normalization needs to be performed on the input feature vector, that is, each dimension of the original vector of the digital type is converted into a decimal number between 0 and 1. Correspondingly, the server can perform normalization processing on the sample feature information of any positive sample or negative sample to obtain a normalized feature, and then the normalized feature is input into the logistic regression model to be trained. The normalization formula is shown in formula (4).

x′＝(x-μ)/σ (4)；

Wherein x' represents the feature vector after normalization, x represents the feature vector before normalization, μ represents the mean, and σ represents the standard deviation.

In an alternative implementation, the server can map the sample feature information of any positive sample or negative sample into a one-hot coded vector, and input the one-hot coded vector into a logistic regression model to be trained. The normalization process can be performed by mapping to a one-hot coded vector.

For example, features related to the number of days and times in the sample feature information, such as the number of days of last week landing on the game, the number of days of last two weeks landing on the game, the number of different machines used in three weeks, the total number of days of last week using the contestants, the total number of days of last two weeks using the contestants, the number of times of purchasing days greater than 31 in all behaviors of purchasing the game acceleration server, and the like are converted into one-hot vectors, and the one-hot vectors do not need to be normalized.

It should be noted that, in order to increase the model training speed, the server can adopt a parallelization mechanism to train the at least two classification models in parallel.

It should be noted that, in the above-mentioned scheme, an optional implementation manner is provided in order to improve the accuracy of the classification model, and correspondingly, the server may further improve the accuracy of a single classification model by adjusting a threshold of the confidence score output by the classification model or adjusting a ratio of positive and negative samples in the training sample, which is not limited in the embodiment of the present application. Optionally, in the model training process, the server may further determine the optimal hyper-parameter by using a grid search or the like, and may also determine the target user by using a multi-model voting method. Optionally, the server may further use interpretable models such as SHAP (SHAPLey adaptive extensions) and LIME (Local interpretive ml-interpretation extension) to calculate the global importance and the Local feature importance, so as to improve the accuracy of the classification model.

It should be noted that, after obtaining the at least two classification models, the server can classify the first user based on the at least two classification models, and determine the probability that the first user redeems the target service again after the target service expires, so as to determine whether to push the target information to the user. Accordingly, the server classifies the user and pushes the offer information, see steps 304 to 307.

304. The server acquires a first user, wherein the first user is a user who exchanges the target service.

In the embodiment of the application, when the target service redeemed by any user is about to expire, the user is the first user.

Optionally, the server may obtain, as the first user, a user that meets a target condition, where the target condition is that the target service expires after the first duration, and the duration of the target service that has been exchanged is not less than the second duration.

For example, if the first duration is 3 days and the second duration is 31 days, the game acceleration service expires after 3 days, and the user who purchased the game acceleration service for 31 days or more last time is the first user.

It should be noted that, because users exchange target services every day, and exchange durations are different, and accordingly, due times of the target services exchanged by the users are not completely the same, the server can acquire different first users every day. Correspondingly, in order to ensure timeliness of the classification model, the server can retrain the classification model at regular intervals, such as every other week, every other three days, every day, and the like, which is not limited by the embodiment of the present application.

305. The server acquires feature information of a first user, wherein the feature information is used for indicating the use feature and the exchange feature of the first user on the target service.

In the embodiment of the application, for a user who redeems a target service, the server can authorize to acquire behavior information and attribute information of the user for using the target service and purchasing the target service, and then the server can acquire the use characteristics of the user for using the target service and the redemption characteristics for redeeming the target service based on the acquired behavior information and attribute information. Optionally, the server may obtain at least two of an acceleration time characteristic, an acceleration delay packet loss characteristic, a login characteristic, an acceleration content characteristic, a service switching behavior characteristic for indicating use of another acceleration service, a first redemption behavior characteristic for indicating recent redemption of an acceleration service, and a second redemption behavior characteristic for indicating historical redemption of an acceleration service of the first user who has redeemed an acceleration service, and then perform data preprocessing on the obtained at least two characteristics to obtain the characteristic information of the first user. The manner in which the server obtains the characteristic information refers to the manner in which the server obtains the sample characteristic information in step 302, and is not described herein again.

306. And the server maps the characteristic information to a target vector space to obtain a characteristic vector.

In the embodiment of the application, the server can perform data processing on the feature information in various ways to obtain the feature vector of the classification model input by the user. The target vector space is the vector space of the one-hot coded vector, and the server maps the characteristic information into the one-hot coded vector and then takes the one-hot coded vector as the characteristic vector. Optionally, the server may further perform normalization processing on the feature information to obtain the feature vector. The embodiment of the present application does not limit this.

307. And the server inputs the feature vectors into at least two classification models respectively, the at least two classification models classify the first user, and at least two classification results are output, wherein the classification results are used for indicating the probability that the first user is a target user, and the target user is a user who cannot exchange target service again.

In this embodiment of the application, the server may be configured to predict the feature information of the first user based on the trained at least two classification models, and each classification model outputs one classification result, so that at least two classification results may be obtained. The classification result output by each classification model can indicate the probability that the first user is the target user, and the higher the probability is, the probability that the first user cannot exchange the target service again after the target service is expired is represented; the lower the probability, the lower the probability that the first user will not redeem the target service again after the target service has expired.

308. And in response to the at least one classification result indicating that the probability that the first user is the target user is greater than the target probability, the server pushes target information to the first user, wherein the target information is used for indicating the purchase of the preferential content of the target service.

In this embodiment of the application, the server may obtain three classification results according to the three models obtained by the training, and then take an intersection of the three classification results, that is, when all three classification results indicate that the probability that the first user is the target user is greater than the target probability, determine that the first user is the target user. Optionally, the server may further determine that the first user is the target user when any classification result indicates that the probability that the first user is the target user is greater than the target probability; or the server can also determine that the first user is the target user when the classification result not less than the target proportion indicates that the probability that the first user is the target user is greater than the target probability. After the server determines that the first user is the target user, the server can push target information for indicating the exchange of the preferential content of the target service to the first user, so that the first user can exchange the target service again after receiving the target information, thereby avoiding the loss of users and improving the total number of users exchanging the target service.

It should be noted that the server can check the predicted effect of the at least two classification models through a comparison test. The server can randomly divide the determined target user into two parts, wherein one part is used as an experimental group, the other part is used as a control group, and the server sends target information to the experimental group to obtain the total income of the experimental group and the total income of the control group. The total income of the comparison group is obviously higher than that of the comparison group, so that the total income can be obviously improved by determining the target user through the classification model to release the coupon.

It should be noted that the server can push different target information to different users, and the different target information includes coupons with different amounts.

In an optional implementation manner, the server can perform weighted summation on at least two probabilities output by at least two classification models to obtain a probability weight, where the probability weight is used to indicate the benefit strength of the benefit content indicated by the target information, and the probability weight is proportional to the benefit strength. That is, the server can determine the amount of the coupon based on the probability weight.

In an alternative implementation manner, the server can issue a large amount of coupons by taking the highest probability of at least two probabilities output by at least two classification models as a probability weight, so that the possibility of user renewal can be improved; of course, the server can also issue a small amount of coupons by taking the lowest probability of the at least two probabilities output by the at least two classification models as a probability weight, so that the overall income can be increased.

Optionally, the server is provided with a basic quota, and the server takes the product of the basic quota and the probability weight as the actual preferential quota indicated in the target information. For example, if the basic benefit is minus 10 yuan and the probability weight is 0.75, the actual benefit amount is 7.5 yuan.

In an optional implementation manner, the server can further set a plurality of probability weight intervals, and one probability weight interval corresponds to one coupon. And the server determines the preferential content indicated by the target information according to the interval of the probability weight.

For example, [0,0.5) corresponds to a coupon of 9, [0.5,0.75) corresponds to a coupon of 8, [0.75,0.85) corresponds to a coupon of 7, [0.85,0.9) corresponds to a coupon of 6, [0.9,0.95) corresponds to a coupon of 5, and [0.95,1) corresponds to a coupon of 4.

It should be noted that the foregoing solution is an optional implementation manner of the information pushing method provided by the present application, and correspondingly, the information pushing method also has other optional implementation manners, such as predicting a renewal user who will redeem the target service again after the expiration by using at least two classification models, and then issuing a coupon to a non-renewal user. Experiments show that the server can randomly divide the predicted renewal user into two parts, one part is used as an experimental group, the other part is used as a control group, and the server issues coupons to the control group to obtain the total income of the experimental group and the total income of the control group. The total income of the experimental group is obviously higher than that of the control group, so that the coupon is not released by predicting the renewal user through the model, and the total income can be obviously improved. In addition, since the proportion of the users who renew within one week after the service is expired is large, these users are also the renewing users. Alternatively, the coupon can be provided with an expiration date, such as 3 days, 5 days, 7 days, and so forth.

In an optional implementation manner, the information push method provided by the embodiment of the present application can be applied to an information push system. Referring to fig. 4, fig. 4 is a schematic structural diagram of an information push system according to an embodiment of the present application. As shown in fig. 4, the information push system is composed of five parts, namely training sample acquisition and feature extraction, model training, prediction sample acquisition and feature extraction, model prediction, and coupon delivery. Wherein, the training sample and the prediction sample are both obtained from the hive database and are updated regularly every day. The training sample feature extraction and the prediction sample feature extraction are to extract features through a unified feature extraction engine, and the model training and the model prediction are finished through the unified model engine. And selecting users needing to release the coupons in the coupon release step, and providing the users to the coupon accurate release system for release.

It should be noted that, in order to make the flow described in the above step 301 to step 308 clearer, refer to fig. 5, where fig. 5 is a flowchart of another information pushing method provided according to an embodiment of the present application. As shown in fig. 5, 501, training sample acquisition; 502. extracting the characteristics of the training samples; 503. training a model; 504. obtaining a prediction sample; 505. extracting the characteristics of the prediction sample; 506. model prediction; 507. and (4) pushing information, namely seven steps. Wherein, step 501 comprises: 5011. acquiring users which expire n days before each day; 5012. dividing positive sample users and negative sample users according to whether the due users are renewed or not after the due users are expired; 5013. and extracting the behavior and attribute data of the positive and negative sample users from the data table. Step 502 includes: 5021. processing the original data of the positive and negative samples into a feature vector to obtain feature information; 5022. and taking feature information of the current day and the past days as training data of the current-day classification model. Step 503 comprises: 5031. training an XGboost model; 5032. training an RF model; 5033. the LR model was trained. Step 504 includes: 5041. acquiring users which expire 3 days ago at present as predicted users; 5042. behavior and attribute data of the predicted user are extracted from the data table. Step 505 comprises: 5051. and processing the original data of the predicted user into a feature vector to obtain feature information. Step 506 includes: 5061. integrating the three models; 5062. taking the intersection of the model outputs as a predicted target user; step 507 includes: 5071. and issuing the coupons to the target users.

In the embodiment of the application, the characteristics beneficial to classifying the first user can be obtained by obtaining the use characteristics and the exchange characteristics of the first user exchanging the target service to the target service, then the first user is classified based on at least two classification models respectively, whether the first user is the target user who cannot continuously exchange the target service can be determined based on at least two obtained classification results, and then target information including preferential content is pushed to the first user determined as the target user, so that the target information can prompt the first user to continuously exchange the target service, the total number of users exchanging the target service is increased, and the loss of users is avoided.

Fig. 6 is a block diagram of an information pushing apparatus provided according to an embodiment of the present application. The apparatus is used for executing the steps executed by the information pushing method, referring to fig. 6, and the apparatus includes: an information acquisition module 601, an information processing module 602, a classification module 603, and an information push module 604.

An information obtaining module 601, configured to obtain feature information of a first user who redeems a target service, where the feature information is used to indicate a usage feature and a redemption feature of the target service for the first user;

an information processing module 602, configured to map the feature information to a target vector space to obtain a feature vector;

a classification module 603, configured to input the feature vector into at least two classification models respectively, classify the first user by the at least two classification models, and output at least two classification results, where the classification results are used to indicate a probability that the first user is a target user, and the target user is a user who will not redeem the target service again;

and the information pushing module 604 is configured to, in response to that the at least one classification result indicates that the probability that the first user is the target user is greater than the target probability, push target information to the first user, where the target information is used to indicate that the offer content of the target service is redeemed.

In an optional implementation manner, the classification module 603 is configured to input the feature vectors into at least two classification models respectively, classify the first user by the at least two classification models, and output the at least two classification results.

acquiring sample characteristic information of the positive sample and the negative sample, wherein the sample characteristic information is used for indicating the use characteristic and the exchange characteristic of the positive sample and the negative sample on the target service;

according to the target date, acquiring a plurality of users of which the target service is redeemed for not less than a fourth time and expires on the target date;

selecting users meeting a first condition from the plurality of users as the positive sample, wherein the first condition is that the target service with not less than a fifth time length is exchanged within a target time range corresponding to the target date;

and selecting users meeting a second condition from the plurality of users as the negative sample, wherein the second condition is that the target service is not redeemed within a target time range corresponding to the target date.

in response to the logistic regression model satisfying an end-of-training condition, treating the logistic regression model as one of the at least two classification models.

In an optional implementation manner, the inputting the sample feature information of any positive sample or negative sample into the logistic regression model corresponding to the iterative process includes:

and normalizing the sample characteristic information of any positive sample or negative sample to obtain normalized characteristics, and inputting the normalized characteristics into the logistic regression model corresponding to the iteration process.

and mapping the sample characteristic information of any positive sample or negative sample into a one-hot coded vector, and inputting the one-hot coded vector into the logistic regression model corresponding to the iteration process.

the information obtaining module 601 is configured to obtain at least two of an acceleration time characteristic, an acceleration delay packet loss characteristic, a login characteristic, an acceleration content characteristic, a service switching behavior characteristic indicating that other acceleration services are used, a first redemption behavior characteristic indicating that the acceleration service is redeemed recently, and a second redemption behavior characteristic indicating that the acceleration service is redeemed historically of the first user redeemed for the target service; and carrying out data preprocessing on the acquired at least two characteristics to obtain the characteristic information of the first user.

In an alternative implementation, the information pushing module 604 is configured to determine at least two probabilities according to the at least two classification results; responding to any probability of the at least two probabilities being larger than the target probability, pushing the target information to the first user; or, in response to the at least two probabilities being both greater than the target probability, pushing the target information to the first user; or, in response to the probability exceeding the target ratio of the at least two probabilities being greater than the target probability, pushing the target information to the first user.

In an optional implementation, the apparatus further includes:

a user obtaining module 605, configured to obtain, as the first user, a user meeting a target condition that the target service expires after the first duration, and a duration of time for which the target service has been purchased is not less than a second duration.

In an alternative implementation, the information processing module 602 is configured to map the feature information into a one-hot coded vector, and use the one-hot coded vector as the feature vector.

It should be noted that: in the information pushing apparatus provided in the foregoing embodiment, only the division of the function modules is illustrated when information is pushed, and in practical applications, the function distribution may be completed by different function modules according to needs, that is, the internal structure of the apparatus is divided into different function modules to complete all or part of the functions described above. In addition, the information pushing apparatus and the information pushing method provided by the above embodiments belong to the same concept, and specific implementation processes thereof are detailed in the method embodiments and are not described herein again.

In this embodiment of the present application, the computer device can be configured as a terminal or a server, when the computer device is configured as a terminal, the terminal can be used as an execution subject to implement the technical solution provided in the embodiment of the present application, when the computer device is configured as a server, the server can be used as an execution subject to implement the technical solution provided in the embodiment of the present application, or the technical solution provided in the present application can be implemented through interaction between the terminal and the server, which is not limited in this embodiment of the present application.

Fig. 7 is a block diagram of a terminal 700 according to an embodiment of the present application. The terminal 700 may be a portable mobile terminal such as: a smart phone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III, motion video Experts compression standard Audio Layer 3), an MP4 player (Moving Picture Experts Group Audio Layer IV, motion video Experts compression standard Audio Layer 4), a notebook computer, or a desktop computer. Terminal 700 may also be referred to by other names such as user equipment, portable terminal, laptop terminal, desktop terminal, and so on.

In general, terminal 700 includes: a processor 701 and a memory 702.

The processor 701 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and so on. The processor 701 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array). The processor 701 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 701 may be integrated with a GPU (Graphics Processing Unit) which is responsible for rendering and drawing the content required to be displayed by the display screen. In some embodiments, the processor 701 may further include an AI (Artificial Intelligence) processor for processing computing operations related to machine learning.

Memory 702 may include one or more computer-readable storage media, which may be non-transitory. Memory 702 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in the memory 702 is used to store at least one program code for execution by the processor 701 to implement the information pushing method provided by the method embodiments herein.

In some embodiments, the terminal 700 may further optionally include: a peripheral interface 703 and at least one peripheral. The processor 701, the memory 702, and the peripheral interface 703 may be connected by buses or signal lines. Various peripheral devices may be connected to peripheral interface 703 via a bus, signal line, or circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 704, a display screen 705, a camera assembly 706, an audio circuit 707, a positioning component 708, and a power source 709.

The peripheral interface 703 may be used to connect at least one peripheral related to I/O (Input/Output) to the processor 701 and the memory 702. In some embodiments, processor 701, memory 702, and peripheral interface 703 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 701, the memory 702, and the peripheral interface 703 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

The Radio Frequency circuit 704 is used for receiving and transmitting RF (Radio Frequency) signals, also called electromagnetic signals. The radio frequency circuitry 704 communicates with communication networks and other communication devices via electromagnetic signals. The rf circuit 704 converts an electrical signal into an electromagnetic signal to transmit, or converts a received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit 704 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and so forth. The radio frequency circuitry 704 may communicate with other terminals via at least one wireless communication protocol. The wireless communication protocols include, but are not limited to: the world wide web, metropolitan area networks, intranets, generations of mobile communication networks (2G, 3G, 4G, and 5G), Wireless local area networks, and/or WiFi (Wireless Fidelity) networks. In some embodiments, the radio frequency circuit 704 may also include NFC (Near Field Communication) related circuits, which are not limited in this application.

The display screen 705 is used to display a UI (User Interface). The UI may include graphics, text, icons, video, and any combination thereof. When the display screen 705 is a touch display screen, the display screen 705 also has the ability to capture touch signals on or over the surface of the display screen 705. The touch signal may be input to the processor 701 as a control signal for processing. At this point, the display 705 may also be used to provide virtual buttons and/or a virtual keyboard, also referred to as soft buttons and/or a soft keyboard. In some embodiments, the display 705 may be one, disposed on a front panel of the terminal 700; in other embodiments, the display 705 can be at least two, respectively disposed on different surfaces of the terminal 700 or in a folded design; in other embodiments, the display 705 may be a flexible display disposed on a curved surface or on a folded surface of the terminal 700. Even more, the display 705 may be arranged in a non-rectangular irregular pattern, i.e. a shaped screen. The Display 705 may be made of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), or the like.

The camera assembly 706 is used to capture images or video. Optionally, camera assembly 706 includes a front camera and a rear camera. Generally, a front camera is disposed at a front panel of the terminal, and a rear camera is disposed at a rear surface of the terminal. In some embodiments, the number of the rear cameras is at least two, and each rear camera is any one of a main camera, a depth-of-field camera, a wide-angle camera and a telephoto camera, so that the main camera and the depth-of-field camera are fused to realize a background blurring function, and the main camera and the wide-angle camera are fused to realize panoramic shooting and VR (Virtual Reality) shooting functions or other fusion shooting functions. In some embodiments, camera assembly 706 may also include a flash. The flash lamp can be a monochrome temperature flash lamp or a bicolor temperature flash lamp. The double-color-temperature flash lamp is a combination of a warm-light flash lamp and a cold-light flash lamp, and can be used for light compensation at different color temperatures.

The audio circuitry 707 may include a microphone and a speaker. The microphone is used for collecting sound waves of a user and the environment, converting the sound waves into electric signals, and inputting the electric signals to the processor 701 for processing or inputting the electric signals to the radio frequency circuit 704 to realize voice communication. For the purpose of stereo sound collection or noise reduction, a plurality of microphones may be provided at different portions of the terminal 700. The microphone may also be an array microphone or an omni-directional pick-up microphone. The speaker is used to convert electrical signals from the processor 701 or the radio frequency circuit 704 into sound waves. The loudspeaker can be a traditional film loudspeaker or a piezoelectric ceramic loudspeaker. When the speaker is a piezoelectric ceramic speaker, the speaker can be used for purposes such as converting an electric signal into a sound wave audible to a human being, or converting an electric signal into a sound wave inaudible to a human being to measure a distance. In some embodiments, the audio circuitry 707 may also include a headphone jack.

The positioning component 708 is used to locate the current geographic Location of the terminal 700 for navigation or LBS (Location Based Service). The Positioning component 708 can be a Positioning component based on the Global Positioning System (GPS) in the united states, the beidou System in china, or the galileo System in russia.

Power supply 709 is provided to supply power to various components of terminal 700. The power source 709 may be alternating current, direct current, disposable batteries, or rechargeable batteries. When the power source 709 includes a rechargeable battery, the rechargeable battery may be a wired rechargeable battery or a wireless rechargeable battery. The wired rechargeable battery is a battery charged through a wired line, and the wireless rechargeable battery is a battery charged through a wireless coil. The rechargeable battery may also be used to support fast charge technology.

In some embodiments, terminal 700 also includes one or more sensors 710. The one or more sensors 710 include, but are not limited to: acceleration sensor 711, gyro sensor 712, pressure sensor 713, fingerprint sensor 714, optical sensor 715, and proximity sensor 716.

The acceleration sensor 711 can detect the magnitude of acceleration in three coordinate axes of a coordinate system established with the terminal 700. For example, the acceleration sensor 711 may be used to detect components of the gravitational acceleration in three coordinate axes. The processor 701 may control the display screen 705 to display the user interface in a landscape view or a portrait view according to the gravitational acceleration signal collected by the acceleration sensor 711. The acceleration sensor 711 may also be used for acquisition of motion data of a game or a user.

The gyro sensor 712 may detect a body direction and a rotation angle of the terminal 700, and the gyro sensor 712 may cooperate with the acceleration sensor 711 to acquire a 3D motion of the terminal 700 by the user. From the data collected by the gyro sensor 712, the processor 701 may implement the following functions: motion sensing (such as changing the UI according to a user's tilting operation), image stabilization at the time of photographing, game control, and inertial navigation.

Pressure sensors 713 may be disposed on a side frame of terminal 700 and/or underneath display 705. When the pressure sensor 713 is disposed on a side frame of the terminal 700, a user's grip signal on the terminal 700 may be detected, and the processor 701 performs right-left hand recognition or shortcut operation according to the grip signal collected by the pressure sensor 713. When the pressure sensor 713 is disposed at a lower layer of the display screen 705, the processor 701 controls the operability control on the UI interface according to the pressure operation of the user on the display screen 705. The operability control comprises at least one of a button control, a scroll bar control, an icon control and a menu control.

The fingerprint sensor 714 is used for collecting a fingerprint of a user, and the processor 701 identifies the identity of the user according to the fingerprint collected by the fingerprint sensor 714, or the fingerprint sensor 714 identifies the identity of the user according to the collected fingerprint. When the user identity is identified as a trusted identity, the processor 701 authorizes the user to perform relevant sensitive operations, including unlocking a screen, viewing encrypted information, downloading software, paying, changing settings, and the like. The fingerprint sensor 714 may be disposed on the front, back, or side of the terminal 700. When a physical button or a vendor Logo is provided on the terminal 700, the fingerprint sensor 714 may be integrated with the physical button or the vendor Logo.

The optical sensor 715 is used to collect the ambient light intensity. In one embodiment, the processor 701 may control the display brightness of the display screen 705 based on the ambient light intensity collected by the optical sensor 715. Specifically, when the ambient light intensity is high, the display brightness of the display screen 705 is increased; when the ambient light intensity is low, the display brightness of the display screen 705 is adjusted down. In another embodiment, processor 701 may also dynamically adjust the shooting parameters of camera assembly 706 based on the ambient light intensity collected by optical sensor 715.

A proximity sensor 716, also referred to as a distance sensor, is typically disposed on a front panel of the terminal 700. The proximity sensor 716 is used to collect the distance between the user and the front surface of the terminal 700. In one embodiment, when the proximity sensor 716 detects that the distance between the user and the front surface of the terminal 700 gradually decreases, the processor 701 controls the display 705 to switch from the bright screen state to the dark screen state; when the proximity sensor 716 detects that the distance between the user and the front surface of the terminal 700 is gradually increased, the processor 701 controls the display 705 to switch from the breath-screen state to the bright-screen state.

Those skilled in the art will appreciate that the configuration shown in fig. 7 is not intended to be limiting of terminal 700 and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components may be used.

Fig. 8 is a schematic structural diagram of a server according to an embodiment of the present application, where the server 800 may generate a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 801 and one or more memories 802, where the memory 802 stores at least one program code, and the at least one program code is loaded and executed by the processors 801 to implement the information pushing method provided by the method embodiments. Of course, the server may also have components such as a wired or wireless network interface, a keyboard, and an input/output interface, so as to perform input/output, and the server may also include other components for implementing the functions of the device, which are not described herein again.

The embodiment of the present application further provides a computer-readable storage medium, which is applied to a computer device, and at least one program code is stored in the computer-readable storage medium, and is loaded and executed by a processor to implement the operations performed by the computer device in the information push method according to the above embodiment.

Embodiments of the present application also provide a computer program product or a computer program comprising computer program code stored in a computer readable storage medium. The processor of the computer device reads the computer program code from the computer-readable storage medium, and executes the computer program code, so that the computer device executes the information pushing method provided in the above-mentioned various alternative implementations.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An information pushing method, characterized in that the method comprises:

2. The method of claim 1, wherein the step of training the at least two classification models comprises:

3. The method of claim 2, wherein the obtaining positive and negative samples comprises:

4. The method of claim 2, wherein the training the at least two classification models based on the sample feature information comprises:

5. The method according to claim 4, wherein the inputting the sample feature information of any positive sample or negative sample into the logistic regression model corresponding to the iterative process comprises:

6. The method according to claim 2, wherein the inputting the sample feature information of any positive sample or negative sample into the logistic regression model corresponding to the iterative process comprises:

7. The method of claim 1, wherein the target service is an acceleration service;

the acquiring of the feature information of the first user redeemed the target service includes:

acquiring at least two of an acceleration time characteristic, an acceleration delay packet loss characteristic, a login characteristic, an acceleration content characteristic, a service switching behavior characteristic used for indicating to use other acceleration services, a first exchange behavior characteristic used for indicating to exchange the acceleration service recently and a second exchange behavior characteristic used for indicating to exchange the acceleration service historically of a first user who exchanged the target service;

and performing data preprocessing on the acquired at least two characteristics to obtain the characteristic information of the first user.

8. The method of claim 1, wherein the pushing targeted information to the first user in response to the at least one classification result indicating that the probability that the first user is a targeted user is greater than a target probability comprises:

determining at least two probabilities according to the at least two classification results;

in response to any probability of the at least two probabilities being greater than the target probability, pushing the target information to the first user; alternatively, the first and second electrodes may be,

in response to the at least two probabilities being greater than the target probability, pushing the target information to the first user; alternatively, the first and second electrodes may be,

in response to a probability exceeding a target ratio of the at least two probabilities being greater than the target probability, pushing the target information to the first user.

9. The method of claim 1, wherein prior to obtaining the characteristic information of the first user redeeming the target service, the method further comprises:

and acquiring a user meeting a target condition as the first user, wherein the target condition is that the target service expires after a first time length, and the time length for exchanging the target service is not less than a second time length.

10. The method of claim 1, wherein mapping the feature information to a target vector space to obtain a feature vector comprises:

and mapping the characteristic information into a one-hot coded vector, and taking the one-hot coded vector as the characteristic vector.

11. An information pushing apparatus, characterized in that the apparatus comprises:

12. The apparatus of claim 11, wherein the training of the at least two classification models comprises:

13. The apparatus of claim 12, wherein the obtaining positive and negative samples comprises:

14. A computer device, characterized in that the computer device comprises a processor and a memory for storing at least one piece of program code, which is loaded by the processor and executes the information pushing method according to any one of claims 1 to 10.

15. A storage medium for storing at least one program code, the at least one program code being configured to perform the information pushing method according to any one of claims 1 to 10.